AI Report Library

Find a prompt, model or category.

What is the difference between GPT-6.1 Sol and Terra?

PromptWhat is the difference between GPT-6.1 Sol and Terra?

what is the difference between 6.1 sol and 6.1 terra

$d-search

Run details and counts
Model
GPT-6.1 Sol
Reasoning
High
Operator
OpenAI
Mode
Codex
Search type
Web Search
Run date
Oct 27, 2026
Pack
d-search
Content language
English
Length
592 words
Links
6
Domains
2
Tables
0
Output
Report contents

GPT-6.1 Sol and Terra

Checked 2026-10-27. OpenAI recommends Sol for complex projects and Terra for concise writing based on supplied material, defined computer-use tasks, and scoped code fixes. It recommends comparing the models on the same inputs and reasoning effort. OpenAI model selection, accessed 2026-10-27.

Sol's model page describes complex coding, computer use, and professional work. GPT-6.1 Sol, accessed 2026-10-27. Terra's page describes a fast, efficient model for visual reasoning, computer use, and writing. GPT-6.1 Terra, accessed 2026-10-27.

At Standard speed, the credit rates per million tokens are 50 input, 2.5 cached input, and 250 output for Sol, and 25, 1.25, and 125 for Terra. Credit rates alone do not determine included subscription usage. ChatGPT pricing, token rates, accessed 2026-10-27.

Sol's API model page lists a 1,050,000-token context window, a 922,000-token maximum input, a 128,000-token maximum output, text and image input, and text output. Its knowledge cutoff is 2026-04-30. GPT-6.1 Sol, accessed 2026-10-27.

Terra's API model page lists the same context, maximum input, maximum output, and modalities. Its knowledge cutoff is 2026-06-01. GPT-6.1 Terra, accessed 2026-10-27. These are API specifications and do not establish effective limits in every Codex workflow.

OpenAI's Work and Codex model guide positions Sol for repeated, long-running work across code, apps, and documents. It positions Terra for everyday writing, visual understanding, and computer use. The guide displays three speed icons for Sol and four for Terra, but gives no measured latency ratio. Work and Codex models, accessed 2026-10-27.

The Terra release note describes image and screen understanding, precise computer use, and clear, direct writing. It announces availability in ChatGPT Work and Codex. What's new, October 26-30, 2026, accessed 2026-10-27.

Recommendation: Start with Sol for research that requires judgment or complex, multi-step work. Try Terra for a defined task with clear inputs and an easily checked result. This is an inference from OpenAI's model-selection guidance, not a measured comparison of the user's tasks.

The published Standard credit rates make Terra half the per-token credit cost of Sol for equal input, cached input, and output token counts. This calculation does not establish that a complete task costs half as much. Models may use different token counts, and included subscription usage follows separate limits. ChatGPT pricing, token rates, accessed 2026-10-27.

Evidence passages

The following excerpts support the task comparison. They are short extracts from the official pages checked on 2026-10-27.

  • Model selection, "When to consider GPT-6.1 Terra": "concise writing based on supplied material" and "scoped code fixes". The following paragraph asks users to compare Terra with Sol using the same inputs and reasoning effort. Model selection, accessed 2026-10-27.
  • Sol model page, introduction: "Near-Astra performance for complex work at a lower cost." GPT-6.1 Sol, accessed 2026-10-27.
  • Terra model page, introduction: "A fast, efficient model for visual reasoning, precise computer use, and clear, natural writing." GPT-6.1 Terra, accessed 2026-10-27.
  • Pricing, "Token rates": Sol's row is 50, 2.5, and 250 credits per million tokens. Terra's row is 25, 1.25, and 125. The section says credit prices alone do not determine included subscription usage. Token rates, accessed 2026-10-27.

No controlled benchmark of the user's tasks was conducted. The focused official-source check supports a task-based choice and the credit-rate comparison. It does not establish a numerical quality or completion-time difference.